Contact Resting - Quadratic Programming

Math and Physics Programming

Started by Jungle Boy May 25, 2005 06:30 PM

4 comments, last by Jungle Boy 18 years, 10 months ago

122

Author

May 25, 2005 06:30 PM

hi guys I'm looking for a fast QP solver for my physic's implementation with friction (for the contact resting, specifically). Is there a "standard algorithm" for that? Thx

dblack

150

May 25, 2005 06:39 PM

Hi,

You probably want to go with an iterative approach(at least to start off with). A good description of iterative solutions to the contact problem(eg SOR etc) is:

http://www.diku.dk/forskning/image/publications/erleben.html

Especially: http://www.diku.dk/~kenny/thesis.pdf

Alternativly there are more accurate direct methods, but they tend to run in polynomial time. (search for dantzig LCP etc). One site to check out is:

http://ioe.engin.umich.edu/people/fac/books/murty/linear_complementarity_webbook/

Which discusses a multitude of methods for solving LCPs(the subclass of quadratic programs needed for contact resolution).

David

Jungle Boy

122

Author

May 25, 2005 11:52 PM

thx for reply David :)

But i'm totally ignorant about that and not very concerned about "how to solve quadratic programming" right now.

I googled several times for LCP implemented algorithms but i didn't find :/ Do you know where i can find an implemented of LCP algorithm?

John Schultz

811

May 26, 2005 12:13 AM

Quote:Original post by Jungle Boy
thx for reply David :)

But i'm totally ignorant about that and not very concerned about "how to solve quadratic programming" right now.

I googled several times for LCP implemented algorithms but i didn't find :/ Do you know where i can find an implemented of LCP algorithm?

ODE implements LCP, and is a complete physics engine. It's free and open source, see www.ode.org. Minimally, examining the source will give you an idea of what it takes to develop a physics engine. There are other free physics engines, though it's not clear if they are also open source.

Amazing Curves Racing